Sundar Pichai

Kevin Roose says Sundar Pichai does not say AGI. Demis Hassabis does, and Sergey Brin does. Roose does not think the labs feel it equally. Anthropic is the most AGI-pilled, OpenAI is a close second, and Google has always struggled. Ilya Sutskever used to walk floor to floor at OpenAI telling people to feel the AGI.

Inside Google, researchers fought it out on internal message boards. One wrote a memo called Say AGI: if Google wants to win in AI, recruit top technical talent, and be careful about how it develops AI, it has to admit, by using the term, that the work is more than tools that make Gmail, Docs, and Google Photos better. It is building toward human-level general intelligence. Someo

Sundar Pichai says Waymo just crossed half a million paid rides per week. The team is focused on scaling that number. A Toronto launch is coming, and 10 new cities are on the plan for this year.

Alex Heath asked whether Waymo keeps partnerships such as Uber, runs a fully owned fleet, or someday sells cars a person could rent out. Sundar Pichai says the space is far from zero-sum and still early. He expects many applications for autonomous driving, and he expects Waymo to work in a myriad of ways, including with other partners. He expects those partners to do well as companies on their own.

Over the next 12 months he expects more of that scaling: international markets, other weather, and hi

Sundar Pichai says Project Suncatcher is Google's pilot, with Planet Labs and SpaceX, for AI data centers in space. Alex Heath asked whether it was still up. Sundar Pichai says it is doing well, he watched the launch, and he expects a chance to say hello to the TPUs.

He compares the bet to starting Waymo. He remembers flying to the desert for the Stanley car in the DARPA challenge, when today's Waymo would have sounded like science fiction. A 10 to 20 year span is a long time, he says, and a world-class team taking the early steps of a deep commitment will hit milestones.

The reason to go now is energy. Sun-synchronous orbits give access to power 24 hours a day. Earth has limited physical

Sundar Pichai says the resolve to keep investing comes from the strength and durability of demand, from consumer products through to enterprise. Alex Heath said Google went cash-flow negative over the summer for the first time since the IPO, and Sundar Pichai confirmed it was free cash flow.

He says a bet of this size needs confidence at several levels. Google is seeing robust demand across that range, and it has its own demand signal from how Google uses AI. Separately, internal evals show the capability curve, including how new models compare with older ones. Asked whether Gemini 5.0 will be much better than Argon, he says yes. Google is also signing durable long-term contracts with commi

Sundar Pichai says Google changed who runs frontier models at DeepMind because that work has become an all-consuming job of its own. It is moving, it is picking up pace, and it runs from data to compute to the algorithms to the teams and the model agents.

Jeff Dean spent 27 years at Google. Sundar Pichai says he is proud of that work, Jeff Dean wanted to go build something he was excited about with a set of people, and Google is excited to see what he does with Discovery Loop.

Demis Hassabis has been thinking about this frontier moment for a long time. Sundar Pichai says Demis Hassabis wants more of his time on getting through it safely and on applying AI to other problems, the work behind

Sundar Pichai says Gemini 4 Argon drew almost 25 trillion tokens in the first 72 hours after launch, about 3 times what Gemini 3.0 saw in that same window. The launch is capacity constrained. Customer feedback has been really good, and he still wants the model to get better at coding. Google's engineers have loved using it internally, and the team keeps shipping refinements from that use.

Alex Heath asked how Google gates a model this large and expensive to serve. Sundar Pichai says Google distilled the Gemini 3.0 models into versions that are nearly as capable and much more efficient, and that work is underway for Gemini 4 so more people can use it. At launch, Google opened Argon to all Ul

Sundar Pichai says he genuinely sees practical commercial applications for quantum in a five-year time frame. If he were sitting in a chair five years out, the Google Cloud team would be negotiating deals on quantum deployments at some scale, the way Google provides TPU pods today.

He frames robustly scaling up quantum systems, with a view toward commercial applications, as almost a 10-year time frame. The future possibilities of quantum are both vast and, in some ways, uncertain. Practical commercial applications inside five years are the ones he says Google would get excited by.

Sundar Pichai says quantum is another technology Google has been committed to for a long time, and he expects

Sundar Pichai says agents will keep using browsers and computers because a person would do the same thing, and because that makes it easier to keep the user in the loop. Agent protocols will develop at the same time.

Alex Heath asked whether the web needs to change so agents can finish tasks, and whether Google is in a position to do something bigger. Sundar Pichai calls the moment complex. Agents working at scale means many parts of the economy have to interoperate in a way that makes sense for all the players. He says the technology frontier will be ahead of the ecosystem. Momentum is already behind MCP, other agent protocols, and payment protocols like UCP, which Google did with a set of

Sundar Pichai says the big compute decisions are still made by people in the loop, including him. He already runs internal agents that brief him on the model development arc and on how compute gets used.

He says catching the last 24 hours across pre-training chat groups, mid-training chat groups, and post-training data could easily take an hour a day. Agents now summarize that flow and bring it to him, and he still goes deeper often. Separate agents watch telemetry on compute usage by product area, efficiency, and utilization, and give him inputs and insights. Alex Heath asks whether any near-term future has AI making the compute decisions. Sundar Pichai says that even five years ago they u

Sundar Pichai says Google is definitely at a stage where AI researchers co-work with the models in a deep way. A lot of humans stay in the loop because of the safety aspects.

The models are involved and maybe suggesting recipes, and they automatically evaluate next checkpoints. He says there is a healthy amount of models and researchers co-working. Some aspects of the pipeline are probably closer to recursive self-improvement, some are less so, and the whole continuum is shifting. Pichai expects the next 12 months to see continued acceleration because of recursive self-improvement.

He says it is important to think about how Google develops the frontier while moving through that shift.

Sundar Pichai says Google partnered with Anthropic because, very early, the founding team looked world-class and had been involved in some of the earliest GPT models. Google could tell quickly that they would be at the frontier. The habit is to identify companies at the cutting edge of technology and partner when the chance is there. SpaceX is the other example he names. Some of those partnerships also took shape around Google's TPUs.

Alex Kantrowitz puts Google's commitment at tens of billions of dollars invested. Pichai does not discuss that figure. He describes the decision as recognizing a frontier team and taking the chance to partner.

Sundar Pichai says Google's cloud business is already giving up short-term revenue, and that the industry is doing some form of the same trade. At any moment the cloud business forgoes that revenue because Google makes compute trade-offs to support frontier AI development. He says that choice is already in place. Alex Kantrowitz frames it as giving researchers the biggest computer even when it hurts today's cloud profits, and he recalls Pichai once calling AI as profound as fire and electricity.

Pichai does not treat the trade as a blank check. Constraints push creativity and more efficient algorithms, including ways to use less memory. What Google is committed to is investing what is neede

Sundar Pichai does not treat this build as a fiber-style overbuild. Over a long enough timeframe it makes complete sense to him. He calls these the early stages of the most consequential technology ever. The compute is in use, not sitting dark, Google is capacity constrained every day, and the applications and the technology are improving quickly.

He plans Google's own compute from demand signals, revenue already committed in contracts, internal evals, how the company uses the models, and where the models are headed. The stance he describes is confident and aggressive, and still disciplined. He does not try to score every other company's risk. An overbuild is possible, he says, because a da

Sundar Pichai says Gemini 4 Argon is one of the really good frontier models. Google took longer on it than on other models, tested it safely, and deployed it carefully. He puts it across agentic work, knowledge work, coding, and creative tasks. In the first 72 hours after launch, use was almost up to 25 trillion tokens, about 3 times what Gemini 3.0 did in its first 72 hours, and Google is capacity constrained while serving it.

The frontier stays competitive and keeps moving. A model can look better across the board than one from two months earlier, and coding still has pockets where the feedback says the work is not done. Google's plan is to keep evolving Argon rather than treat the launch

Sundar Pichai says he holds a deep conviction that improving technology and improving access to technology makes people's lives better, drawn from his own life. The optimism is not a belief that you just build whatever you want. You work hard to harness technology so it works well for society and to build it in the service of humanity, making the human condition better. He treats that goal as the defining aim, and within that framework he is very optimistic.

He thinks the world is in very early stages of the progress still coming across many areas. Healthcare is one area people already discuss, and a lot of the progress is still ahead. He also points to making many things fundamentally chea

Sundar Pichai says data centers in space should make sense for the rise in compute Google expects to need. A sun-synchronous orbit would give 24/7 access to the most abundant energy available, which he says fits that growth. The project is two to three years old. He compares it with bets that looked like science fiction and later became real: quantum, about 10 years in, and Waymo, almost 20 years after the DARPA Grand Challenge car Stanley.

The work is still hard, and the team uses disciplined milestones. A recent one was sending TPUs to space. During this conversation he was waiting on a hello-world signal from them. He says Google will keep investing.

Sundar Pichai says Google's unified Gemini agent is a persistent agent the company announced for enterprise and is bringing to consumers. The agents run 24/7 in the cloud. Each query effectively spins up a micro VM. The runtime, the harness, and the model form a headless service that works everywhere and keeps context with the person.

People can address the agent in Workspace and in Gemini. Agents are meant to drive a unified customer experience. A broad product portfolio that moves fast will move faster in some areas than in others, and the uneven pace counts as valid feedback. Google is putting a lot of time and effort into the agentic experience so unification and capability work across

Sundar Pichai says Google's ground rule is to invest in enough compute to stay at the AI frontier, and that priority is non-negotiable. In 2015 he told the company to be AI first, so AI runs across search, cloud, and Waymo. Google DeepMind gets cutting-edge compute access for research and frontier model development before other plans.

Each business works from its own long-range plan. The cloud team brings projections, demand signals, and committed contracts, and those signals shape a long-term compute path. Google builds its own compute and its own data centers. Lead times are long, so the compute on hand today comes from decisions he made in 2024. Google and other companies are short, and

Sundar Pichai says frontier AI models matter because they will be better at developing the next models in an era of recursive self-improvement, and being at that frontier also puts a company at the innovation frontier.

Google saw a steep inflection after the newest model reached its developers. It handles problems earlier models could not. Models of every size still have a place, which is why Google built Flash and the other Flash lines. A strong product does not always need the most advanced model, because the top models tend to be slower and more expensive. Google still cares about the full parade of frontier models and is committed to doing all of it well.

A build that took months of wo